IEEE Communications Magazine • November 201542 0163-6804/15/$25.00 © 2015 IEEE

Ian F. Akyildiz is with theGeorgia Institute of Tech-nology.

Pu Wang is with WichitaState University.

Zhi Sun is with the StateUniversity of New York atBuffalo.

ABSTRACT

The majority of the work on underwater com-munication has mainly been based on acousticcommunication. Acoustic communication facesmany known problems, such as high propagationdelays, very low data rates, and highly environ-ment-dependent channel behavior. In this article,to address these shortcomings, magnetic induc-tion is introduced as a possible communicationparadigm for underwater applications. Accord-ingly, all research challenges in this regard areexplained. Fundamentally different from the con-ventional underwater communication paradigm,which relies on EM, acoustic, or optical waves,the underwater MI communications rely on thetime varying magnetic field to covey informationbetween the transmitting and receiving parties.MI-based underwater communications exhibitseveral unique and promising features such asnegligible signal propagation delay, predictableand constant channel behavior, sufficiently longcommunication range with high bandwidth, aswell as silent and stealth underwater operations.To fully utilize the promising features of under-water MI-based communications, this articleintroduces the fundamentals of underwater MIcommunications, including the MI channel mod-els, MI networking protocols design, and MI-based underwater localization.

INTRODUCTIONUnderwater communication networks have drawnthe attention of the research community in the lastdecade and a half, driven by a wealth of theoreti-cal and practical challenges. This growing interestcan largely be attributed to new applicationsenabled by large-scale networks of underwaterdevices (e.g., underwater static sensors, unmannedautonomous vehicles, and autonomous robots),which are capable of harvesting information fromthe aquatic and marine environment, performingsimple processing on the extracted data, and trans-mitting it to remote locations. Significant results inthis area over the last few years have ushered in asurge of civil and military applications. However,the underwater environment imposes great chal-lenges on reliable and real-time communicationsprimarily based on acoustic waves as well as elec-tromagnetic (EM) and optical waves.

EM waves experience high attenuation underwater, which severely limits the achievable com-munication range. To increase the communicationrange, large antennas are required for low-fre-quency EM communication, which is not practicalfor small underwater vehicles and robots. Forexample, the antenna size of an EM transmitter isa couple of meters for a 50 MHz operating fre-quency. Optical waves experience multiple scatter-ings of light, which results in intersymbolinterference and short transmission range [1]. Toprolong the transmission range, the transmissionof optical signals requires high precision in point-ing narrow laser beams at the receiver, which isdifficult for highly mobile underwater vehicles androbots. The common and most used acousticwaves, while promising long communicationranges under water, exhibit high propagation delayalong with unreliable and unpredictable channelbehavior and low data rate, which is caused bycomplex multi-path fading, prevalent Dopplereffects, and significant variation of these proper-ties due to temperature, salinity, or pressure [2].

In the last decade our research on magneticinduction (MI)-based communications in chal-lenged and harsh environments [3, 4] such assoil, oil reservoirs, and water pipelines hasdemonstrated many promising features of MI-based communication. Adopting similar commu-nication principles as in [3, 4], recent researchprovides mathematical analysis of the MI-basedcommunication channel in underwater applica-tions [5–7]. Compared to commonly used acous-tic, optical, and EM communication, MI-basedcommunication has the following advantages:

Negligible signal propagation delay: Differentfrom acoustic waves that propagate at a speed of1500 m/s under water, MI waves propagate at aspeed of 3.33 × 107 m/s under water. Thisextremely high propagation speed of MI wavescan significantly improve the delay performanceof underwater communications, while facilitatingthe design and implementation of the underwaternetworking protocols, such as medium accesscontrol (MAC) and routing, and the underwaternetworking services (e.g., localization). More-over, physical layer synchronization among wire-less devices becomes simple and reliable due tothe negligible delay and stable channel.

Predictable and constant channel response:Since the radiation resistance of a coil is much

UNDERWATER WIRELESS COMMUNICATIONS ANDNETWORKS: THEORY AND APPLICATION

Ian F. Akyildiz, Pu Wang, and Zhi Sun

Realizing Underwater Communicationthrough Magnetic Induction

WANG2_LAYOUT.qxp_Author Layout 10/30/15 3:15 PM Page 42

IEEE Communications Magazine • November 2015 43

smaller than that of an electric dipole, only avery small portion of energy is radiated to the farfield by the coil. Hence, compared to acousticcommunication, multi-path fading is not an issuefor MI-based underwater communication. More-over, because of the high propagation speed ofMI waves, the frequency offsets caused byDoppler effect can be greatly mitigated. Withoutsuffering from multi-path fading and Dopplereffect, the MI channel conditions (e.g., data rateand packet loss rate over a given transmissionrange) are highly constant and predictable. More-over, without suffering from light scattering as inoptical communications, the transmission rangeand channel quality of MI communications areindependent of water quality factors such aswater turbidity. In addition, both acoustic andoptical communications have to deal with a highlevel of acoustic and ambient light noises. TheEM noise experienced by MI channels is limitedunder water because the high-frequency noise isabsorbed by the water medium.

Sufficiently large communication range withhigh data rate: In MI-based communications,the transmission and reception are accomplishedthrough the use of a pair of small-size wire coils,that is, coil antennas. Different from the dipoleantenna used in most EM wave-based communi-cations, there is no minimum frequency belowwhich the antenna cannot work. On the onehand, the time varying magnetic field can begenerated no matter how small the coil is at theMI transmitter. On the other hand, as long asthere is magnetic flux going through the coil, theMI receiver can capture the signal even if thefrequency is as low as the Megahertz band. Thisproperty means that each small coil antenna canbe utilized to emit low-frequency MI signals,which allow small underwater robots and vehi-cles to communicate over sufficiently long dis-tances. Moreover, the operating frequency of MIcoils can reach Megahertz bands while maintain-ing predictable and constant channel quality,which leads to much higher data rates than inthe acoustic communications.

Stealth underwater operations: While acous-tic and optical communications depend on thegeneration, propagation, and reception of audi-ble sounds or visible lights, respectively, under-water MI communications utilize non-audibleand non-visible MI waves, suitable for a widerange of civilian and military applications thatrequire stealth underwater operations.

The comparison of MI, EM, acoustic, andoptical communications is summarized in Table1. It is also worth noting that the cost of MI coilsis very low, generally less than US$1.00 per coilfor mass production. Therefore, it is very suitablefor large-scale deployment of MI-based underwa-ter nodes. The promising features of underwaterMI communications can enable many new andemerging underwater applications:

Collaborative sensing and tracking withunderwater swarming robots: Fish behaviorshows an astonishing ability to efficiently findfood sources and favorable habitat regionsthrough schooling behavior, that is, rapid orient-ing and synchronized moving of a group of fishwith respect to environmental gradients, such aslocal variations in chemical stimuli such as odor-ant plumes or other environmental propertiessuch as phytoplankton density. Underwater MIcommunications can enable a swarm of underwa-ter robots (e.g., agile robotic fish) to mimic thiscollective and synchronized intelligence of fish byexchanging control and environmental gradientinformation with guaranteed delay bounds. Insuch a way, swarming robots can collaborativelytrack sources of pollution, toxicity, and biohazardwith high convergence speed and accuracy.

Stealth and real-time underwater surveillanceand patrol: The high bandwidth along with theconstant and reliable channel conditions achievedby underwater MI communications can enablereal-time underwater surveillance, which demandshigh-speed delivery of a large volume of multime-dia contents (e.g., audio, video, and scalar data).In addition, the stealth and silent features ofunderwater MI communications allow underwatersurveillance to be carried out in stealth mode.

Disaster assessment, search, and rescue in acluttered underwater environment: Underwaterstructures, such as leaking submarine cabins,sunken ships, and completely submerged build-ings, can create a cluttered underwater environ-ment, which is constituted by confined,hard-to-reach, and complex underwater spaces.Such environments prevent the deployment ofbulky underwater vehicles and the application ofconventional acoustic communications, whichinevitably experience severe multi-path fadingwith significantly degraded channel quality. Inthis case, a small and agile underwater robotequipped with small coil antenna can maneuverin such clustered environments much more easi-ly, while allowing reliable and real-time MI

Table 1. Comparison of underwater MI, EM, acoustic, and optical communications.

Communicationparadigm

Propagationspeed

Datarates

Communicationranges Channel dependency Stealth

operation

MI 3.33 × 107 m/s ~ Mb/s 10–100 m Conductivity Yes

EM 3.33 × 107 m/s ~ Mb/s £ 10 m Conductivity, multipath Yes

Acoustic 1500 m/s ~ kb/s ~ km Multipath, Doppler, temperature, pressure,salinity, environmental sound noise Audible

Optical 3.33 × 107 m/s ~ Mb/s 10–100 m Light scattering, line of sight communication,ambient light noise Visible

WANG2_LAYOUT.qxp_Author Layout 10/30/15 3:15 PM Page 43

IEEE Communications Magazine • November 201544

underwater communications for comprehensivein situ disaster assessment as well as timely sur-vivor search and rescue.

In the remainder of this article, we first focuson the MI channel modeling and physical layertechniques. Then we introduce the design guide-lines for underwater MI networking protocols.Finally, we highlight the underlying principles ofthe MI-based underwater localization method.

UNDERWATERMI CHANNEL MODELING

MI CHANNEL PATH LOSSWith MI communication, the data information iscarried by a time varying magnetic field. Such amagnetic field is generated by the modulated sinu-soid current along an MI coil antenna at the trans-mitter. The receiver retrieves the information bydemodulating the induced current along the receiv-ing coil antenna, as shown in Fig. 1. In MI commu-nications, the transmission distance is shorter thana wavelength. As a result, the communicationchannel experiences less absorption due to thelossy underwater medium. Moreover, the multi-path fading is negligible in underwater MI systems.As previously mentioned, the MI transceivers workat the Megahertz band that has a wavelength oftens of meters. Since the communication range inunderwater MI systems is within one wavelength,even if there are multiple paths between thetransceivers, the phase shifting of multiple paths isso small that the coherence bandwidth is muchlarger than the system bandwidth. Hence, the fad-ing and channel distortion are negligible.

To rigorously characterize the above uniqueunderwater MI communication channel, an ana-lytical channel model is of great importance. Inthe MI channel models [3], the EM fields aroundthe transmitter and receiver coils are firstexpressed using Maxwell’s equations. Based onthe field analysis, the coupling between thetransceiver coils is modeled using the equivalentcircuit method. Finally, the MI path loss can bederived as a function of the operating frequency,

transmission distance, coil antenna size, numberof turns, the relative angle between the twocoils, and the underwater environmental condi-tions, especially the water conductivity.

While the MI channel models in [3, 5, 6] consid-er the directional coil antennas, Fig. 2 shows thenumerical path loss of the underwater MI commu-nication channel with small-size omnidirectionalcoil antennas [7]. We consider the MI transceiversto be equipped with coil antennas with 10 cm radiusand 20 turns of AWG26 wire. The operating fre-quency is 10 MHz. Three types of underwater envi-ronments are considered: sea water withconductivity 4 S/m, lake water with conductivity0.005 S/m, and drinking water with conductivity0.0005 S/m. According to Fig. 2a, using a pocket-sized wireless device can achieve 20 m communica-tion range in the drinking water, more than 10 mrange in the lake water, but less than 1 m range inthe sea water. (Note that here the maximum com-munication range is considered to be the distancewhere the path loss reaches 100 dB.) Such big dif-ferences in communication ranges are due to theorders of magnitudes differences in the mediumconductivities in the sea water, lake water, anddrinking water. Highly conductive sea water inducessignificant Eddy current that incurs very high pathloss, which is the problem for both the EM wavesand MI techniques in the seawater applications.

The transmission range of underwater MI com-munications can be increased by using the optimaloperating frequency and larger coil antennas. Fig-ures 2b and 2c show the influence of the operatingfrequency and coil antenna size on the MI channelpath loss in the lake water. As the operating fre-quency increases from 100 kHz to 15 MHz, theMI path loss decreases at first and then keepsincreasing after a turning point. The minimumvalue in Fig. 2b indicates that there is an optimaloperating frequency for MI communications in thelake water. On the one hand, the higher frequencycan enhance the MI coupling between the trans-mitter and the receiver. On the other hand, due tothe non-zero connectivity in the lake water, thehigher frequency also encounters higher loss dueto Eddy current. Hence, there is an optimal fre-quency that achieves the minimum path loss andmaximizes the communication range. Moreover,the coil size has a monotonically positive influenceon the MI coupling, which means that less pathloss and longer communication distance can berealized by increasing the coil antenna size, asshown in Fig. 2c. In addition, the path loss can befurther reduced by increasing the wire turns of thecoil antenna. For example, based on our channelmodel, the path loss can be reduced by 6 dB if thenumber of turns is doubled. It should be notedthat the performance shown in Fig. 2 is based onthe assumption that the transmitter coil andreceiver coil are coaxially positioned. However,the orientation of the coil antenna can be highlyrandom, since underwater robots and vehicles canmove and rotate freely in the target underwaterenvironment. Hence, we introduce underwaterMI-based omnidirectional communication.

OMNIDIRECTIONAL COMMUNICATIONDespite the promising properties of MI under-water communications, according to the derivedchannel models, if the underwater robots and

Figure 1. Illustration of MI-based communications with a tridirectional coilantenna.

Transmitter

XTX

XTX

ZTX

yTX

yRX

XRX

ZRX

XRXDirectionalMI coil:

OmnidirectionalMI coil:

Receiver

WANG2_LAYOUT.qxp_Author Layout 10/30/15 3:15 PM Page 44

IEEE Communications Magazine • November 2015 45

vehicles use the single coil antenna, the path lossperformance of the MI channel varies accordingto the uncontrollable coil orientations. To solvethe problem, we propose utilizing a tridirectionalcoil antenna at the receiver and transmitter,which consists of three coils that are perpendicu-lar to each other, as shown in Fig. 1. Here weshow that underwater robots and a vehicleequipped with a tridirectional coil antenna con-stitute an omnidirectional receiver, and the coildirection has almost no effect on communicationperformance. It should be noted that, for illus-trative purposes, we only let one of the threecoils work at the transmitter in Fig. 1. Due tothe field distribution pattern of the coil, theorthogonal coils on the same wireless device donot interfere with each other since the magneticflux generated by one coil becomes zero at theother two orthogonal coils.

To prove the advantages of the tridirectionalcoil antenna, we numerically compare the per-formance of the systems using either a unidirec-tional or tridirectional coil in the MIcommunication channel. When the receiver is aunidirectional coil, the path losses with differentantenna orientations are shown in Fig. 3a. Incontrast, when the receiver is a tridirectionalcoil, the path losses with different antenna orien-tations are given in Fig. 3b. According to theresults, the tridirectional coil antenna has muchlower path loss and more reliable performancethan the unidirectional coil when the antennadeviates from its optimal orientation. Moreover,since the path loss of the tridirectional coilantenna does not obviously vary when the orien-tation of the antenna changes, the tridirectionalreceiver can be regarded as omnidirectional.

OPEN RESEARCH ISSUES3D MI channel modeling: Although tri-coil

MI communications have demonstrated promis-ing omnidirectional features, an analytical 3Dmodel of the tri-coil MI channel still needs to bedeveloped. Based on such an analytical channelmodel, the optimal omnidirectional performancecan be achieved by properly combining thereceived signals at the three coil antenna ele-ments by maximizing transmission gain throughoptimal power allocation, and by adopting spa-tial-temporal coding mechanisms.

Transmission range extension through under-water MI waveguide: The transmission range ofMI communications can be extended by adopt-ing the MI waveguide technique. The channelmodel of the MI waveguide was developed in [3]delicately for underground wireless communica-tions. The MI waveguide consists of a series ofrelay coils between two transceivers. An MIrelay is just a simple coil without any powersources or processing devices. Since the MItransceivers and relays are coupled, the relayswill get the induced currents one by one up tothe receiver, so the signal strength at the receiv-er side becomes large enough over long distance.Since the original channel model of MI-waveg-uide is developed for 2D directional MI-coilantennas [3], the omnidirectional MI-waveguidechannel model still needs to be developed in the3D aquatic space.

Underwater MI coil antenna design: The

parameters of the coil antenna, including thesize, number of turns, as well as the lumped andthe distributed impedance, have significant influ-ence on the underwater MI communications.Therefore, to maximize the underwater MI com-munication distance as well as the bandwidth,the optimal coil antenna parameters need to bedesigned by jointly considering the trade-offbetween the induced current and the absorptiondue to skin depth, the trade-off between the MIpath loss and the antenna size, as well as the

Figure 2. Path loss of the underwater MI communication channel: a) pathloss of MI underwater communications; b) influence of operating fre-quency (lake water); c) influence of coil antenna size (lake water).

0 2 4 6 8 10 12 14

(a)

(b)

(c)

16 18 200

20

40

60

80

100

120

140

160

Transmission distance (m)

Path

loss

(dB

)

Drinking waterLake waterSeawater

05

1015

x 106

0

5

10

15

200

50

100

150

200

Operating frequency (Hz)

Transmission distance (m)

Path

loss

(dB

)

00.1

0.20.3

0.40.5

05

1015

200

50

100

150

200

Coil radius (m)

Transmission distance (m)

Path

loss

(dB

)

WANG2_LAYOUT.qxp_Author Layout 10/30/15 3:15 PM Page 45

IEEE Communications Magazine • November 201546

trade-off between the quality factor of the coilcircuit and the system bandwidth.

Underwater MI transceiver design: To realizethe underwater MI communication scheme, theMI transceiver needs to be designed and imple-mented, which consists of the RF front-end, ana-log-to-digital/digital-to-analog converter(ADC/DAC), modulator, and equalizer. In par-ticular, since the MI channel is insensitive tomultipath fading, Doppler effect, and underwa-ter acoustic noise, an extremely high-order mod-ulation scheme and corresponding channelequalization scheme can be jointly designedbased on the underwater MI channel models,which can provide high data rate MI communi-cation links under water.

Energy modeling: Since the underwaterrobots and vehicles are usually powered by bat-teries and are expected to operate for a longtime period, energy-aware communication, and

networking mechanisms are of great importance,which rely on an accurate energy consumptionmodel of the whole system. Therefore, a preciseenergy model needs to be developed based onthe unique underwater MI channel model aswell as the optimal antenna and transceiverparameters.

UNDERWATER MI-BASEDCROSS-LAYER DESIGN

OVERVIEW OFMI-BASED CROSS-LAYER PROTOCOL

The performance of underwater MI-communica-tions can be further enhanced through the cross-layer design approach. For example, we haveproposed a cross-layer communication frameworkfor the MI-based wireless sensor networks [4].This cross-layer module provides a unified solu-tion for underground, soil, oil reservoir, andunderwater environments because it can adaptive-ly adjust the communication parameters accordingto the channel characteristics such as bandwidthand path loss under different environments.

More specifically, the proposed cross-layercommunication framework is a distributed rout-ing/MAC/PHY solution and allows each node tojointly optimize the next-hop selection, transmit-ted power, modulation scheme, and forward errorcorrection (FEC) coding rate, with the objective ofsimultaneously minimizing energy consumptionand maximizing packet transmitted rate, while sat-isfying the application-specific QoS requirementssuch as delay and packet loss rate. The proposedsolution relies on a highly scalable geographicalrouting paradigm and adopts the direct sequencecode-division multiple access (DS-CDMA) as theMAC-layer protocol, the performance of which isfurther enhanced through distributed game-theo-retic power control to combat the near-far prob-lem [4]. Our preliminary results show that theproposed cross-layer protocol outperforms the lay-ered protocol solutions with 50 percent energysavings and 6 dB throughput gain.

OPEN RESEARCH ISSUESUnderwater MAC protocol design: The exist-

ing MAC design in underwater mainly aims tocombat the adverse impact of acoustic channels[8, 9]. The underwater MAC protocols via MI-based communications are still unexploited. Car-rier sense multiple access (CSMA) schemes areeasily implemented in a distributed and low-com-plexity manner, and are robust to time-varyingnetwork topology caused by node mobility. How-ever, conventional underwater acoustic communi-cations face fundamental limitations inimplementing CSMA-like schemes because ofthe high and variable propagation delay of under-water acoustic waves [1]. On the contrary, thereare no such barriers to implement CSMA-likeschemes for underwater MI communicationsbecause the propagation delay MI waves underwater are negligible. In addition, the relativelyhigh operating frequency of MI waves paves theway for the design and implementation of fre-quency-division multiple access (FDMA) proto-cols under water, which is not feasible forunderwater acoustic communication because of

Figure 3. Comparison between the path loss in the MI system with unidirec-tional and tridirectional coils (as a function of antenna orientations): a) path loss of a unidirectional receiver; and b) path loss of a tridirectionalreceiver.

00.2

0.40.6

0.81

00.5

11.5

240

45

50

55

60

65

70

Relative angle in x p

lane (π )

Relative angle in y plane (π)

Path

loss

(dB

)

00.2

0.40.6

0.81

00.5

11.5

240

45

50

55

60

65

70

Path

loss

(dB

)

(a)

(b)

Relative angle in y plane (π)Relative

angle in x plane (π )

WANG2_LAYOUT.qxp_Author Layout 10/30/15 3:15 PM Page 46

IEEE Communications Magazine • November 2015 47

its low operating frequency. Similar to MI-basedcommunications, EM-based communications alsoexhibit low propagation delay and can operate athigh frequency. However, the existing MAC pro-tocols designed for EM-based communications inthe terrestrial case cannot work effectively andefficiently for underwater MI-based communica-tions because they fail to capture the unique fea-tures of the MI channel. For example, becauseMI-based communications depend on the time-varying near-field magnetic strength, the signalstrength of MI-based communications will dropdramatically in the far field. This means thatinstead of considering the complicated physicalmodel, that is, the signal-to-interference-plus-noise ratio (SINR) model, the MAC protocoldesign for underwater MI-based ad hoc networkscan safely adopt the simple protocol model,where the impact of interference from a trans-mitting node is only determined by whether ornot a receiver resides within the communicationrange of this transmitting node. Moreover,because of the constant and predicable MI chan-nel, each node can accurately estimate its instan-taneous transmission rate only based on itsrelative location to the receiving party. This cangreatly enhance the performance of the celebrat-ed maximum weight scheduling protocols, thethroughput optimality of which depends on thechannel estimation accuracy.

Routing protocol design: Because of the longpropagation delay and high bit error rate ofacoustic channels, it is very difficult to providereliable, timely, and energy-efficient data trans-fer in large-scale underwater networks over mul-tihop communications. To address the challengesof acoustic channels, sophisticated underwaterrouting protocols [10, 11] have been developed.However, the promising features of MI channelsnecessitate revisiting the underwater routingprotocol design, taking into account the 3D fea-ture of the underwater environment and therobust underwater MI channel. Moreover, dif-ferent from the routing operations for EM,acoustic, or optical communications, the nodesin underwater MI-based communications canforward the packets to the next-hop node by act-ing as either a passive relay working in the MIwaveguide mode or a conventional active relayby first receiving the packets and then transmit-ting them to the next-hop node. A passive relaydoes not consume any energy, while the forward-ing distance is relatively small. An active relaycan achieve longer forwarding distance by induc-ing the transmitting and receiving energy con-sumption. Therefore, by adaptively selectingdifferent relay modes, the optimal routing proto-cols can be designed to yield the best trade-offbetween network lifetime and network latency.

Error control techniques: Since the underwa-ter MI channel exhibits totally different charac-teristics from conventional acoustic channels, it isnecessary to investigate the most suitable errorcontrol technique. For example, an MI channelexperiences much lower variations, and such astable channel may only require low-complexityFEC code with high coding rate to achieve highreliability. Our recent research [4] has shown thatthe low-complexity Bose, Ray-Chaudhuri, Hoc-quenghem (BCH) code, which is more energy-

efficient than convolutional codes, is suitable forMI-based communications in challenging under-ground environments. Therefore, it is necessaryto investigate the optimal FEC coding schemesthat yield the best trade-off among the complexi-ty, energy efficiency, and error correction capa-bility of underwater MI-based communications.

UNDERWATERMI-BASED LOCALIZATION

Localization is an important functionality forunderwater communication networks. In particu-lar, coordination and maneuvering of underwa-ter robots and vehicles requires each robot to beaware of the relative 3D positions of its immedi-ate neighbors as well as the absolute coordinatesin the 3D underwater space. However, theabsence of GPS underwater imposes great chal-lenges in the realization of effective relative andabsolute localization under water.

The existing underwater localization methodscan be mainly categorized into acoustic-based anddead-reckoning ones, both of which have funda-mental limitations. Acoustic-based methods, suchas long baseline (LBL) and short baseline (SBL)systems, determine the relative or absolute posi-tions of underwater vehicles by measuring the trav-el time of acoustic waves between vehicles orbetween the vehicle and anchor nodes, based onthe measured or assumed sound speed, and/or theknown locations of the anchor nodes. The acous-tic-based methods require large acoustic antennasfor emitting low-frequency sound, and are thus notsuitable for small-scale underwater robots, and suf-fer unavoidable localization errors caused by theinherent multi-path propagation of acoustic wavesespecially in cluttered, confined, or shallow under-water environments. On the other hand, dead-reckoning methods estimate the current position ofan underwater vehicle based on its last knownposition along with its speed and heading mea-sured by the onboard sensors, such as gyroscopes,Doppler velocity sonar (DVS), and accelerometers.Utilizing the dead-reckoning methods, the errorsin the position estimate grow without bounds asthe underwater vehicle travels continuously.

To conquer the above problems, we proposethe novel received magnetic field strength(RMFS)-based localization method, which usesthe by-product of MI-based communication withtridirectional antenna: the RMFS between the 3× 3 transmitting and receiving coil antennas.The novelty of the proposed solution relies onthe unique multi-path fading free propagationproperties of MI-based signals and the inherentorthogonality of tri-coil MI antennas, whichguarantee accurate, simple, and convenient rela-tive localization strategy. Moreover, such amethod does not induce additional cost, occupyextra space, or increase the weight of theresource constrained underwater nodes. Specifi-cally, each time underwater nodes communicatewith each other, the RMFS is measured. Thenthe mutual distance between underwater nodescan be estimated using the MI channel modeland maximum likelihood estimation (MLE)method. Because of the multi-path fading freefeature, the RMFS-based localization method

This cross-layer

module provides a

unified solution for

underground, soil,

oil reservoir, and

underwater environ-

ments because it can

adaptively adjust the

communication

parameters accord-

ing to the channel

characteristics such

as bandwidth and

path loss under dif-

ferent environments.

WANG2_LAYOUT.qxp_Author Layout 10/30/15 3:15 PM Page 47

IEEE Communications Magazine • November 201548

will yield much more accurate estimation thanthe conventional acoustic-based solutions. Rely-ing on real-time RMFS measurement, it doesnot induce accumulated estimation error as thedead-reckoning methods do. Moreover, since theMI-based technique is sensitive to direction, byusing three coils in orthogonal planes, we canestimate the distance in each of the three direc-tions separately, and calculate the distance andangular coordinates of the other underwaternodes relative to a fixed known reference [12].The open research issues include the develop-ment of robust localization under inter-node MIinterference for swarming underwater nodes.

CONCLUSIONSIn this article, we introduce a new underwatercommunication paradigm, namely underwatermagnetic-induction (MI) communication. Fun-damentally different from the conventionalunderwater communication paradigm, whichrelies on EM, acoustic, or optical waves, under-water MI communications rely on the time-vary-ing magnetic field to covey information betweenthe transmitting and receiving parties. MI-basedunderwater communications exhibit severalunique and promising features such as negligiblesignal propagation delay, predictable and con-stant channel response, sufficiently large com-munication range with high bandwidth, andsilent and covert underwater operations. To fullyutilize the promising features of underwater MI-based communications, this article introducesthe fundamentals of underwater MI communica-tions, including the MI channel models, MI net-working protocol design, and MI-basedunderwater localization.

ACKNOWLEDGMENTThis work was supported by the US NationalScience Foundation (NSF) under Grant No.CNS-1446557 and CNS-1446484. We thank Tom-maso Melodia, Dario Pompili, Ozgur Akan,Mari Carmen Domingo, and Eylem Ekici fortheir valuable comments.

REFERENCES[1] I. Akyildiz, D. Pompili, and T. Melodia, “Underwater Acous-

tic Sensor Networks: Research Challenges,” Elsevier Ad HocNetworks J., vol. 3, no. 3, Feb. 2005, pp. 257–79.

[2] L. Lanbo, Z. Shengli, and C. Jun-Hong, “Prospects andPproblems of Wireless Communication for UnderwaterSensor Networks,” Wireless Commun. and Mobile Com-puting, vol. 8, no. 8, 2008, pp. 977–94.

[3] Z. Sun and I. F. Akyildiz, “Magnetic Induction Commu-nications for Wireless Underground Sensor Networks,”IEEE Trans. Antenna and Propagation, vol. 58, no. 7,2010, pp. 2426–35.

[4] S. C. Lin et al., “Distributed Cross-Layer Protocol Designfor Magnetic Induction Communication in WirelessUnderground Sensor Networks,” IEEE Trans. WirelessCommun., vol. 14, no. 7, 2015, pp. 4006–19.

[5] M. C. Domingo, “Magnetic Induction for UnderwaterWireless Communication Networks,” IEEE Trans. Anten-nas and Propagation, vol. 60, no. 6, 2012, pp. 2929–39.

[6] B. Gulbahar and O. B. Akan, “A Communication Theo-retical Modeling and Analysis of Underwater Magneto-Inductive Wireless Channels,” IEEE Trans. WirelessCommun., vol. 11, no. 9, 2012, pp. 3326–34.

[7] H. Guo, Z. Sun, and P. Wang, “Channel Modeling of MIUnderwater Communication Using Tri-Directional CoilAntenna,” Proc. IEEE GLOBECOM ’15, San Diego, CA,2015.

[8] S. Shahabudeen, M. Motani, and M. Chitre, “Analysis ofa High Performance MAC Protocol for UnderwaterAcoustic Networks,” IEEE J. Oceanic Eng., vol. 39, no.1, 2014, pp. 74–89.

[9] Y. B. Zhu et al., “Toward Practical MAC Design forUnderwater Acoustic Networks,” IEEE INFOCOM, 2013,pp. 683–91.

[10] Z. Zhou and J. H. Cui, “Energy Efficient Multi-PathCommunication for Time-Critical Applications in Under-water Sensor Networks,” ACM MohiHoc, 2008.

[11] A. Jain and X. Zhang, “Aqua-Route: Reliable and Ener-gy-Efficient Routing Protocol in Underwater WirelessSensor Networks,” ACM WUWNET, 2012.

[12] X. Tian, Z. Sun, and P. Wang, “On Localization forMagnetic Induction-Based Wireless Sensor Networks inPipeline Environments,” IEEE ICC, 2015.

BIOGRAPHIESIAN F. AKYILDIZ [F’96] (ian@ece.gatech.edu)received his B.S.,M.S., and Ph.D. degrees in computer engineering from theUniversity of Erlangen-Nuremberg, Germany, in 1978, 1981and 1984, respectively. Currently, he is the Ken Byers ChairProfessor in Telecommunications with the School of Electricaland Computer Engineering, Georgia Institute of Technology(Georgia Tech), Atlanta, director of the Broadband WirelessNetworking (BWN) Laboratory, and chair of the Telecommu-nication Group at Georgia Tech. Since 2013, he has been aFiDiPro Professor (Finland Distinguished Professor Programsupported by the Academy of Science) with the Departmentof Electronics and Communications Engineering, TampereUniversity of Technology, Finland, and the founding directorof the Nano Communications Center. Since 2008, he has alsobeen an honorary professor with the School of ElectricalEngineering at Universitat Politecnica de Catalunya, Barcelona,Spain, and the founding director of the NaNoNetworkingCenter in Catalunya. Since 2011, he has been a consultingchair professor with the Department of Information Technol-ogy, King Abdulaziz University, Jeddah, Saudi Arabia. He isEditor-in-Chief of Elsevier’s Computer Networks Journal andfounding Editor-in-Chief of Elsevier’s Ad Hoc Networks Jour-nal, Elsevier’s Physical Communication Journal, and Elsevier’s .He is an ACM Fellow (1997). He has received numerousawards from IEEE and ACM. His current research interests arein wireless sensor networks in challenged environments suchas underwater and underground, nanonetworks, terahertzband, 5G cellular systems, molecular communication, andsoftware defined networking.

PU WANG [M] (pu.wang@wichita.edu) received his B.E.degree in electrical engineering from Beijing Institute ofTechnology, China, in 2003, and his M.E. degree in electri-cal and computer engineering from Memorial University ofNewfoundland, Canada, in 2008. He received his Ph.D.degree in electrical and computer engineering from Geor-gia Tech under the guidance of Prof. Ian F. Akyildiz inAugust 2013. Currently, he is an assistant professor withthe Department of Electrical Engineering and ComputerScience at Wichita State University, Kansas. He received theBWN Lab Researcher of the Year Award from Georgia Techin 2012. He received the TPC top ranked paper award atIEEE Dy SPAN 2011. He was also named Fa ellow of theSchool of Graduate Studies, Memorial University of New-foundland in 2008. His current research interests are wire-less sensor networks, cognitive radio networks, cloud radioaccess networks, software defined networking, cyber-aquatic systems, nanonetworks, and wireless communica-tions in challenged environments.

ZHI SUN [M] (zhisun@buffalo.edu) received his B.S. degreein telecommunication engineering from Beijing Universityof Posts and Telecommunications and his M.S. degree inelectronic engineering from Tsinghua University, Beijing,China, in 2004 and 2007, respectively. He received hisPh.D. degree in electrical and computer engineering fromGeorgia Tech under the guidance of Prof. Ian F. Akyildiz in2011. Currently, he is an assistant professor in the Electri-cal Engineering Department at the State University of NewYork at Buffalo. Prior to that, he was a postdoctoral fellowat Georgia Tech. He won the Best Paper Award at IEEEGLOBECOM 2010. He was given the outstanding graduateaward at Tsinghua University in 2007. He received the BWNresearcher of the year award at Georgia Tech in 2009. Hisexpertise and research interests lie in wireless communica-tions, wireless sensor networks, and cyber physical systemsin challenged environments.

To fully utilize the

promising features of

underwater

MI-based communi-

cations, this article

introduces the

fundamentals of

underwater MI

communications,

including the MI

channel models,

MI networking

protocol design, and

MI-based underwa-

ter localization.

WANG2_LAYOUT.qxp_Author Layout 10/30/15 3:15 PM Page 48